Vehicle loan risk prediction method and device, storage medium and electronic equipment

Abstract

The invention provides a vehicle loan risk prediction method and device, a storage medium and electronic equipment, and relates to the field of machine learning. The electronic equipment obtains overdue risk characteristics of the borrower in the current repayment period; processing the overdue risk feature through a risk assessment model of a neural network architecture to obtain a first risk score of the borrower in the current repayment period; processing the overdue risk feature through a pre-fitted autoregressive moving average model to obtain a second risk score of the borrower in the current repayment period; and obtaining a comprehensive risk score of the borrower according to the first risk score and the second risk score. Therefore, because the risk assessment model of the neural network architecture can better capture the influence of the sudden events, the defect that the autoregressive moving average model is insensitive to the sudden events is made up, and the prediction precision of the overdue risk is improved.

Description

Technical Field

[0001] This application relates to the field of machine learning, and more specifically, to a vehicle loan risk prediction method, device, storage medium, and electronic device. Background Art

[0002] In modern financial risk management, predicting risks is a crucial task. Traditional risk prediction models are mainly based on historical data and fixed risk characteristics for modeling, and then these models are used to predict future risks. However, this method has some obvious drawbacks. Since this method usually assumes that risk characteristics are static, it ignores the fact that risk characteristics may change over time. Even though some techniques have tried to improve the model by adapting to changes in risk characteristics, these improvement methods are still not sensitive enough to sudden risks. Summary of the Invention

[0003] To overcome at least one deficiency in the prior art, this application provides a vehicle loan risk prediction method, device, storage medium, and electronic device, specifically including:

[0004] In a first aspect, this application provides a vehicle loan risk prediction method, and the method includes:

[0005] Obtain the overdue risk characteristics of the borrower in the current repayment period;

[0006] Process the overdue risk characteristics through a risk assessment model with a neural network architecture to obtain a first risk score of the borrower in the current repayment period;

[0007] Process the overdue risk characteristics through a pre-fitted autoregressive moving average model to obtain a second risk score of the borrower in the current repayment period;

[0008] Obtain a comprehensive risk score of the borrower according to the first risk score and the second risk score.

[0009] In a second aspect, this application provides a vehicle loan risk prediction device, and the device includes:

[0010] A risk characteristic module, configured to obtain the overdue risk characteristics of the borrower in the current repayment period;

[0011] A risk score module, configured to process the overdue risk characteristics through a risk assessment model with a neural network architecture to obtain a first risk score of the borrower in the current repayment period;

[0012] The risk score module is further configured to process the overdue risk characteristics through a pre-fitted autoregressive moving average model to obtain a second risk score of the borrower in the current repayment period;

[0013] A comprehensive risk module, configured to obtain a comprehensive risk score of the borrower according to the first risk score and the second risk score.

[0014] In a third aspect, the present application provides a storage medium storing a computer program, which, when processed by a processor, implements the vehicle loan risk prediction method described above.

[0015] In a fourth aspect, the present application provides an electronic device, which includes a processor and a memory. The memory stores a computer program, which, when processed by the processor, implements the vehicle loan risk prediction method described above.

[0016] Compared with the prior art, the present application has the following beneficial effects:

[0017] The present application provides a vehicle loan risk prediction method, device, storage medium and electronic device. The electronic device obtains the overdue risk characteristics of the borrower in the current repayment period; processes the overdue risk characteristics through a risk assessment model with a neural network architecture to obtain a first risk score of the borrower in the current repayment period; processes the overdue risk characteristics through a pre-fitted autoregressive moving average model to obtain a second risk score of the borrower in the current repayment period; and obtains a comprehensive risk score of the borrower according to the first risk score and the second risk score. In this way, since the risk assessment model with a neural network architecture can better capture the impact of sudden events, it compensates for the defect that the autoregressive moving average model is insensitive to sudden events and improves the prediction accuracy of overdue risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present application and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 It is a schematic flowchart of the vehicle loan risk prediction method provided by the embodiment of the present application;

[0020] Figure 2 It is one of the detailed schematic flowcharts of the vehicle loan risk prediction method provided by the embodiment of the present application;

[0021] Figure 3 It is a schematic diagram of the feature change information provided by the embodiment of the present application;

[0022] Figure 4It is the second detailed flowchart of the vehicle loan risk prediction method provided by the embodiment of the present application;

[0023] Figure 5 It is the third detailed flowchart of the vehicle loan risk prediction method provided by the embodiment of the present application;

[0024] Figure 6 It is the structural schematic diagram of the vehicle loan risk prediction device provided by the embodiment of the present application;

[0025] Figure 7 It is the structural schematic diagram of the electronic device provided by the embodiment of the present application. Detailed implementation manners

[0026] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some but not all of the embodiments of the present application. Usually, the components of the embodiments of the present application described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0027] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application claimed, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0028] It should be noted that: similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0029] In the description of the present application, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance. In addition, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitations, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0030] Based on the above statement, as introduced in the background art, even though some technologies have tried to improve the model by adapting to changes in risk characteristics, these improvement methods are still not sensitive enough to sudden risks.

[0031] Exemplarily, take the Autoregressive Integrated Moving Average Model (ARIMA) as an example (hereinafter simply referred to as the ARIMA model). The ARIMA model is a commonly used time series prediction method that combines the autoregressive model (AR) and the moving average model (MA). The basic idea of the ARIMA model is to analyze the historical data of the time series, identify the patterns and trends therein, and use these patterns to predict future values. To achieve this goal, the ARIMA model consists of three components: the autoregressive term (AR), the differencing term (I), and the moving average term (MA). The autoregressive term represents the linear relationship between the current value and the values in several previous periods. The differencing term is used to make the time series stationary, and the moving average term is used to capture the autocorrelation of the error term. Through the combination of these three parts, the ARIMA model can effectively capture the short-term fluctuations and long-term trends in the time series.

[0032] However, although the ARIMA model performs well in dealing with stationary time series data, it is not sensitive enough to sudden risks. This is mainly because the ARIMA model mainly relies on the patterns and trends in historical data for prediction. When sudden events (such as economic crises, natural disasters, etc.) occur, these events often have no precedents and cannot be predicted through historical data. Therefore, the ARIMA model may not be able to adjust its prediction model in time to cope with these sudden changes, resulting in inaccurate predictions.

[0033] Based on the discovery of the above technical problems, the inventors have proposed the following technical solutions through creative labor to solve or improve the above problems. It should be noted that the defects existing in the above solutions in the prior art are the results obtained by the inventors through practice and careful research. Therefore, the discovery process of the above problems and the solutions proposed by the embodiments of the present application below for the above problems should be the contributions made by the inventors to the present application during the invention creation process, and should not be understood as the technical content known to those skilled in the art.

[0034] In view of the above problems, this embodiment provides a vehicle loan risk prediction method. As Figure 1 shown, the method includes:

[0035] S1, obtaining the overdue risk characteristics of the borrower in the current repayment period.

[0036] S2, processing the overdue risk characteristics through a risk assessment model with a neural network architecture to obtain the first risk score of the borrower in the current repayment period.

[0037] S3. Process the overdue risk characteristics through the pre-fitted autoregressive moving average model to obtain the second risk score of the borrower in the current repayment cycle;

[0038] S4. Obtain the comprehensive risk score of the borrower based on the first risk score and the second risk score.

[0039] In this way, since the risk assessment model with a neural network architecture can better capture the impact of sudden events, it thus makes up for the defect that the autoregressive moving average model is insensitive to sudden events and improves the prediction accuracy of overdue risks.

[0040] It should be noted that for the solutions involved in this application, which involve data related to user privacy, they are explicitly informed and authorized by the user before being collected and generated during the execution of the embodiments of this application. At the same time, for the personal information involved in the embodiments of this application, the setting location of its storage device complies with the laws and regulations requirements of the country / region where the location of the occurrence of the above data-related behaviors is located. The above data-related behaviors include, but are not limited to: authorization, generation, use, storage, etc. It can be understood that for the overdue risk characteristics used in implementing this solution, the setting location of its storage device complies with the laws and regulations requirements of the country / region where the location of the occurrence of the above data-related behaviors is located. The above data-related behaviors include, but are not limited to: authorization, generation, use, storage, etc.

[0041] In addition, it should also be understood that for the vehicle loan risk prediction method of this embodiment, the electronic device implementing this method can be, but is not limited to, a mobile terminal, a tablet computer, a laptop computer, a desktop computer, and a server, etc. The server can be a single server or a server group. The server group can be centralized or distributed (for example, the server can be a distributed system). In some embodiments, the server can be local or remote relative to the user terminal. In some embodiments, the server can be implemented on a cloud platform; only as an example, the cloud platform can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an inter-cloud, a multi-cloud, etc., or any combination thereof. In some embodiments, the server can be implemented on an electronic device with one or more components.

[0042] To make the solution provided in this embodiment clearer, the following takes the server as the electronic device implementing this method to Figure 1Each step in the method shown will be elaborated in detail. However, it should be understood that the operations in the flowchart may not be implemented in sequence, and steps without a logical context relationship can be reversed or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart or remove one or more operations from the flowchart under the guidance of the content of this application. Therefore, continue to refer to Figure 1 , the method includes:

[0043] S1, obtain the overdue risk characteristics of the borrower in the current repayment period.

[0044] In this embodiment, the above-mentioned multiple overdue sub-characteristics include dynamic characteristics and static characteristics that can reflect whether the borrower will be overdue. Exemplarily, the multiple overdue sub-characteristics may include the borrower's credit history characteristics, income stability characteristics, vehicle-related characteristics, economic environment risk characteristics, user personal characteristics, and vehicle positioning information. These characteristics will be described in detail below:

[0045] (1) The borrower's credit history characteristics, including credit card overdraft records, repayment records of other loans, credit scores, overdue repayment records, and default records.

[0046] The credit card overdraft record reflects the borrower's use of the credit card, including whether there is an overdraft behavior and the frequency and amount of the overdraft. If the borrower often overdrafts and fails to repay in a timely manner, this indicates that their financial management ability is poor and there may be a default during the vehicle loan repayment process.

[0047] The repayment records of other loans include the repayment records of all the borrower's previous loans (such as housing loans, consumer loans, etc.). If the borrower has repeatedly overdue or defaulted on loans, this indicates that their repayment willingness is low and the possibility of defaulting in the future during the vehicle loan repayment process is relatively high.

[0048] The credit score is a quantitative indicator used to evaluate the borrower's creditworthiness, usually calculated comprehensively by a credit scoring agency based on the borrower's historical repayment records, overdraft situation, etc. The higher the credit score, the better the borrower's credit and the lower the default risk; conversely, a lower credit score means a higher default risk.

[0049] The default record focuses on whether the borrower has defaulted on previous loans. For example, within two years before applying for a vehicle loan, a certain customer's credit card repayment has been overdue [X] times in total, and the longest consecutive overdue period has reached [X] months. Based on this, the financial institution can initially determine that the customer has a very high default risk during the vehicle loan repayment process.

[0050] Therefore, through a comprehensive evaluation of the borrower's historical repayment records, overdraft situation, credit score, etc., potential default risks can be effectively identified.

[0051] (2) The income stability feature is another important part of the car loan risk assessment in this embodiment, mainly used to examine factors such as the borrower's occupation type, working years, income source, etc., to judge the stability and reliability of their income.

[0052] Occupation type, as there are significant differences in the income stability and reliability among different occupations. For example, occupations such as public servants, state-owned enterprise employees, and formal employees of large enterprises have relatively lower default risks compared to freelancers, self-employed individuals, or small and micro business owners due to the stability of their jobs and the reliability of their income. Such occupations usually have fixed salaries and welfare benefits, reducing the default risk caused by unstable income.

[0053] Working years, borrowers with longer working years usually have higher income stability and job security. In contrast, borrowers with shorter working years may face greater income fluctuations and uncertainties. For example, borrowers with a working year exceeding [X] years have significantly higher income stability than those with a working year less than [X] years.

[0054] Income source, in addition to occupation type and working years, income source is also an important factor in measuring income stability. For example, borrowers with multiple stable income sources (such as fixed salary, rental income, investment income, etc.) have higher income stability compared to borrowers with only a single income source. Diversified income sources can effectively alleviate the risks brought by a single income source.

[0055] Income change trend, by analyzing the income change trend of the borrower in the past few years, the income stability can be further understood. If the borrower's income has shown an upward trend or remained stable in the past few years, it indicates that their income stability and repayment ability are stronger; on the contrary, if the income shows a downward trend, the default risk is higher.

[0056] Occupation prospect, some industries have better occupation prospects and greater income growth potential. For example, high-tech industries, financial industries, etc. These industries usually have good career development prospects and relatively high income levels. Therefore, borrowers engaged in these industries usually have higher income stability and repayment ability.

[0057] Therefore, in this embodiment, by comprehensively evaluating aspects such as the borrower's occupation type, working years, income source, income change trend, and occupation prospect, the income stability and repayment ability of the borrower can be effectively identified.

[0058] (3) Vehicle-related features are an important part of the car loan risk assessment, mainly involving information such as the value, use, and purchase channel of the vehicle, and can evaluate the value and reliability of the vehicle as collateral.

[0059] Vehicle value, the market value of a vehicle is one of the key factors determining the loan amount and loan ratio for vehicle loans. Research has found that the higher the market value of a vehicle, the higher its security as collateral. Financial institutions usually refer to the vehicle valuation reports provided by third-party appraisal agencies to determine the actual value of the vehicle. In addition, the residual value rate of the vehicle is also an important consideration factor. Vehicles with a high residual value rate are more likely to obtain a higher recovery value when disposed of in the future.

[0060] Vehicle use, the use of a vehicle can be divided into two major categories: commercial use and personal use. Vehicles for personal use usually have higher stability and a lower depreciation rate, while commercial vehicles, due to frequent use and a higher degree of wear and tear, have a faster depreciation rate. Therefore, the risk of using a vehicle for personal use as collateral is relatively low, while commercial vehicles require a more cautious assessment.

[0061] Vehicle brand and model, different brands and models of vehicles have different resale value retention capabilities and liquidity in the market. Well-known brands and high-end models usually have a higher market recognition and stronger resale value retention capabilities, while the resale value retention capabilities of some niche brands or low-end models are relatively weak. Therefore, the brand and model of a vehicle are also taken into consideration in vehicle loan assessments.

[0062] Purchase channel, the purchase channel of a vehicle also affects its value and reliability as collateral. Vehicles purchased through regular channels usually have complete procedures and higher transparency, while vehicles purchased through irregular channels may have problems such as incomplete procedures and unclear ownership, increasing the risk of vehicle disposal. Therefore, financial institutions are more inclined to accept vehicles purchased through regular channels as collateral.

[0063] Vehicle mileage, the mileage of a vehicle is an important indicator to measure the usage condition of the vehicle. Vehicles with a lower mileage usually have better vehicle conditions and higher residual values, while vehicles with a higher mileage may have higher maintenance costs and lower residual values. Therefore, the mileage of a vehicle is also one of the key factors to be considered in vehicle loan assessments.

[0064] Vehicle insurance situation, the vehicle insurance situation is also an important factor in assessing its risk. Comprehensive vehicle insurance can not only protect the vehicle from losses in the event of an accident, but also improve the overall safety of the vehicle. Therefore, the vehicle insurance situation report provided by the insurance company is also one of the important bases for financial institutions to assess the risk of vehicle loans.

[0065] Therefore, by comprehensively evaluating the value, usage, brand and model, purchase channel, mileage, and insurance situation of the vehicle, vehicle-related characteristics can effectively identify the risks of using the vehicle as collateral. These characteristics can not only reflect the basic condition of the vehicle but also help financial institutions more accurately predict future repayment risks, thereby taking appropriate measures to manage and reduce the risks of auto loans.

[0066] (4) Economic environment risk is also an important part of auto loan risk assessment, mainly involving the impact of macroeconomic factors and market conditions on borrowers' repayment ability. Specifically, these factors include the following aspects:

[0067] Macroeconomic conditions directly affect borrowers' employment opportunities, income levels, and overall financial conditions. For example, a slowdown in economic growth will lead to an increase in unemployment and a decrease in income, thereby increasing the risk of borrowers' default. On the contrary, strong economic growth is conducive to improving borrowers' income levels and employment stability and reducing the default risk.

[0068] Interest rate changes have a direct impact on borrowers' repayment pressure. When interest rates rise, borrowers' repayment burden increases, which may lead to some borrowers being unable to repay the loan on time, thereby increasing the default risk. Conversely, a decrease in interest rates can relieve borrowers' repayment pressure and improve their repayment ability.

[0069] A high inflation rate will erode borrowers' real income, reduce their purchasing power, and thus affect their repayment ability. A high inflation rate will also make borrowers pay more interest and principal when repaying the loan, increasing their financial pressure. On the contrary, a low inflation rate helps maintain borrowers' purchasing power and income levels and reduces the default risk.

[0070] Regional economic conditions also affect borrowers' repayment ability. There are more employment opportunities and higher income levels in economically developed regions, so borrowers have stronger repayment ability. In economically underdeveloped regions, there are fewer employment opportunities and lower income levels, so borrowers have weaker repayment ability and higher default risk.

[0071] Changes in policies and regulations may also affect borrowers' repayment ability. For example, tax policies and credit policies introduced by the government may directly affect borrowers' income and expenditure. In addition, changes in regulatory policies, such as strengthening credit review and increasing the down payment ratio, will also affect borrowers' repayment ability.

[0072] The market competition situation can also affect the borrower's repayment ability. In a highly competitive market environment, borrowers may face greater economic pressure, thus increasing the default risk. While in a relatively stable market environment, the repayment ability of borrowers is usually more stable.

[0073] Therefore, by comprehensively evaluating aspects such as the macroeconomic situation, interest rate changes, inflation rate, regional economic situation, policy and regulatory changes, and market competition situation, the economic environment risk can effectively identify the repayment risk of borrowers under changing economic environments.

[0074] (5) The personal characteristics of users are a key component in the risk assessment of auto loans, mainly involving factors such as the borrower's personal information, credit record, repayment willingness and behavior, etc., to evaluate their reliability and risk level as borrowers.

[0075] Age and occupation, research has found that the age and occupation of borrowers have a great impact on their repayment ability and stability. Borrowers who are young and at the beginning of their careers may lack stable work experience and a high income level, while borrowers who are older and have a stable occupation usually have a high income and good career prospects. In addition, whether the borrower has a stable occupation, such as civil servants, teachers, doctors, etc., these occupations usually have a high income stability and repayment ability.

[0076] Credit record, the borrower's credit record is an important basis for evaluating their credit status. A good credit record indicates that the borrower has a good repayment history and reputation, while a bad credit record means that the borrower may have had overdue repayments or other credit problems. Financial institutions usually refer to the borrower's credit report to evaluate their credit score and historical repayment behavior.

[0077] Income and debt situation, the income level and debt situation of borrowers are important factors in evaluating their repayment ability. The higher the borrower's income and the lower their debt, the stronger their repayment ability and the lower the default risk. Financial institutions will evaluate the stability of their income and the proportion of debt through materials such as income certificates and bank statements provided by borrowers.

[0078] Asset situation, the asset situation of borrowers is also an important factor in evaluating their repayment ability. Owning fixed assets of high value (such as real estate, stocks, deposits, etc.) can serve as additional repayment guarantees. Financial institutions will understand their financial stability and potential repayment sources by evaluating the asset situation of borrowers.

[0079] Repayment willingness and behavior, the borrower's repayment willingness and historical behavior are also important indicators for assessing their credit risk. Whether the borrower has good repayment habits, such as repaying on time, prepaying, etc., reflects their serious attitude towards debts. In addition, whether the borrower has a record of using other loans or credit cards, and whether these records are good, will also affect their credit score.

[0080] Educational background, the borrower's educational background also reflects to a certain extent their repayment ability and credit quality. Borrowers with higher education usually have higher professional qualities and repayment abilities, while borrowers with lower educational levels may face greater financial pressure and repayment difficulties.

[0081] Family situation, the borrower's family situation, such as the number of family members, marital status, etc., will also affect their repayment ability and stability. Borrowers from single-parent families or those with heavy family burdens may face greater economic pressure, while borrowers with fewer family members, stable marriages, and less economic burden usually have higher repayment abilities.

[0082] Therefore, by comprehensively evaluating the borrower's age and occupation, credit record, income and debt situation, asset status, repayment willingness and behavior, educational background, and family situation, the user's personal characteristics can effectively identify the borrower's credit risk and repayment ability.

[0083] (6) Vehicle location information, through which information such as the vehicle's driving trajectory, speed, mileage, etc. can be obtained, and these information help to more accurately grasp the borrower's vehicle usage situation, detect abnormal behaviors in a timely manner, thereby improving the accuracy of risk prediction.

[0084] As an alternative implementation, in order to inform the risk assessment model which dynamic features are more important and which are relatively less important, the overdue risk features of the current repayment cycle include multiple overdue sub-features and a preset weight assigned to each overdue sub-feature. In this regard, it can be understood that during the risk assessment, the overdue risk features of the borrower in the current repayment cycle will be broken down into multiple specific sub-features, and a pre-set weight will be assigned to each sub-feature. Such a design is to ensure that during the risk assessment process, the impact of different features on the overall risk can be reasonably quantified and reflected.

[0085] Exemplarily, the server can group various dynamic features at different time levels, such as daily level changes, weekly level changes, monthly level changes, and annual level changes; then, encode the non-numerical features in each group for subsequent further processing and calculation. When assigning weights, the server can assign weights to each sub-feature from aspects such as cycle change frequency, intention degree, policy sensitivity, impact degree, economic indicators, etc.

[0086] For the periodic change frequency, the server can analyze the periodic change frequencies of various dynamic features and assign different weights according to the frequency levels. For example, if a certain feature changes every day, it will be given a higher weight; conversely, if the feature changes less frequently, it will be given a lower weight.

[0087] For the intention level, the server can assign different weights according to the influence of features on the intention levels of car purchase, car replacement, loan, etc. within different time periods. For example, in February, users have a greater intention for loans, so the relevant features in this month will be given higher weights.

[0088] For the policy sensitivity, the server can analyze the dynamic features that have a greater impact on policy sensitivity, such as the new energy replacement subsidy policy, and assign different weights according to the sensitivity levels. If a certain policy change has a greater impact on risk features, then the features related to this policy will be given higher weights.

[0089] For the influence degree, the server can consider factors such as the vehicle management, production, and monetary policies of the country in the current year, and assign different weights according to the influence degrees of these factors on dynamic features. For example, if the GDP of a certain region grows rapidly, which will affect the local loan demand and repayment ability, then the features of this region will be given higher weights.

[0090] For economic indicators, the server can weight the affected dynamic features according to factors such as the GDP and per capita income of the country's provinces to reflect the impact of the economic environment on risk features. For example, if the per capita income of a certain province is relatively high, then the loan risk features of this province will be given lower weights.

[0091] Based on the above description of the borrower's risk features in the implementation, the following continues to describe Figure 1 step S2 in

[0092] S2. Process the overdue risk features through the risk assessment model of the neural network architecture to obtain the first risk score of the borrower in the current repayment cycle.

[0093] In this embodiment, the risk assessment model of the above neural network architecture can be a large language model after fine-tuning. Since this model has learned how to understand and process complex text information through a large amount of training data and can capture the correlations and potential patterns between different risk characteristics. Therefore, when the risk characteristics of the above borrower are input into the risk assessment model in text form, the risk model will first parse and understand the input text, identify the key risk characteristics and their weights. Then, the model will use its internal neural network structure and, through a series of calculations and reasoning processes, analyze how these characteristics act together on the overall risk. Finally, the model outputs a comprehensive first risk score, which reflects the model's prediction of the overdue risk of the borrower during the current repayment cycle.

[0094] Therefore, in this way, the fine-tuned large language model can not only efficiently process diverse input data but also provide more accurate and reliable risk assessment results. Moreover, through additional training data and the fine-tuning process, the model has learned how to quickly adapt to and understand newly emerging information or changes. When faced with sudden events, the model can quickly parse the relevant text data and identify the key information and abnormal situations. For example, when conducting risk assessment, if there is a sudden major market fluctuation or important events affecting the borrower's credit status are reported in the news, the fine-tuned large language model can capture this information in a timely manner and accordingly adjust the risk assessment results.

[0095] Exemplarily, the large language model can be Qwen-14B. This model has undergone large-scale deep learning training and has powerful text understanding and generation capabilities. Therefore, after fine-tuning Qwen-14B, it can be used to predict the risk of borrower overdue.

[0096] Based on the description of the first risk score acquisition method in the above embodiment, the following continues to describe Figure 1 step S3 in

[0097] S3, process the overdue risk characteristics through a pre-fitted autoregressive moving average model to obtain the second risk score of the borrower in the current repayment cycle.

[0098] It should be understood that as a commonly used time series analysis tool, the ARMA model can capture the trends and seasonal components in the data and smooth them. In this application, the model will first fit the historical repayment data of the borrower collected to determine the model parameters. Then, using these parameters, the model can process and predict the overdue risk characteristics of the borrower. To achieve this purpose, the ARMA model can identify and quantify the time series characteristics of various risk factors and then calculate the second risk score of the borrower in the current repayment cycle.

[0099] Compared with large language models, the ARMA model relies on numerical inputs to perform calculations and analyses, and thus cannot directly process non-numerical data. To enable the ARMA model to work effectively, the input risk sub-features must be converted into a quantitative form, that is, converted into numerical values by some method. For example, some qualitative features of borrowers (such as credit records, repayment behaviors, etc.) may need to be quantified by scoring or coding. These quantified data can better represent the changing trends and patterns of risk features, enabling the ARMA model to capture the time series characteristics of these features and perform trend analysis and prediction.

[0100] Based on the description of the second risk score acquisition method in the above embodiments, the following continues Figure 1 to illustrate step S4 in

[0101] S4. Obtain the comprehensive risk score of the borrower according to the first risk score and the second risk score.

[0102] In this way, by combining the first risk score generated by the risk assessment model and the second risk score generated by the ARIMA model, the advantages of both models can be comprehensively utilized to improve the accuracy of overall prediction.

[0103] In this embodiment, both the above first risk score and the second risk score are numerical values between 0 and 1. The larger the value, the higher the risk of overdue. Therefore, as an optional implementation manner, the first risk score and the second risk score can be weighted and summed using a preset weight to obtain the comprehensive risk score.

[0104] However, it is found in the practical process that when the difference between the first risk score and the second risk score is large, it means that there is a large disagreement in the recognition results of the two models, which is often caused by the insensitivity of the ARIMA model to sudden information. Therefore, the preset weight cannot adapt to the above situation. For this, as Figure 2 shown, this embodiment provides the following implementation manner of step S4:

[0105] S4-1. Obtain the feature change information between the overdue risk features of the previous repayment cycle and the overdue risk features of the current repayment cycle.

[0106] In this embodiment, the server can subtract each risk sub - feature in the overdue risk features of the previous repayment period from each risk sub - feature in the overdue risk features of the current repayment period to obtain feature change information. If the change value of a certain risk sub - feature is large, it indicates that this change may be sudden information, reflecting new risk factors that the borrower has recently encountered; on the contrary, if the change value is small, it may belong to non - sudden information, that is, the normal fluctuation of the borrower's risk features. In this way, sudden information and non - sudden information can be more accurately distinguished.

[0107] Exemplarily, as Figure 3 shown, the figure shows 5 risk sub - features of the previous repayment period and 5 risk sub - features of the current repayment period. Figure 3 shows 5 risk sub - features of the previous repayment period and 5 risk sub - features of the current repayment period, and their corresponding feature values are shown in the following table:

[0108]

[0109]

[0110] Due to Figure 3 the feature change information in, it is not difficult to see that the change amplitude between risk sub - feature A and risk sub - feature B in two adjacent repayment periods increases, while the change amplitudes of other risk sub - features are smaller. It should be understood that the above examples are only for facilitating the explanation of feature change information. The features and values in the examples will change with the actual sub - features.

[0111] Based on the above description of feature change information, continue to refer to Figure 2 , after step S4 - 1, step S4 further includes:

[0112] S4 - 2, optimizing the feature change information to obtain optimized feature change information.

[0113] Among them, the optimized feature change information enhances the sudden information included in the feature change information and suppresses the non - sudden information in the feature change information. It should be understood that, as Figure 3 shown, for different risk sub - features, the value ranges of their feature values are significantly different. Therefore, they cannot be directly used to measure the degree of event change. In this regard, this embodiment proposes an optimization method that combines the difference between the first risk score and the second risk score and the feature change information to enhance the sudden information in the feature change information and suppress the non - sudden information. As Figure 4 shown, step S4 - 2 may include:

[0114] S4-2-2. Obtain the divergence index between the risk assessment model and the autoregressive moving average model based on the first risk score and the second risk score.

[0115] S4-2-3. Optimize the feature change information using the divergence index to obtain the optimized feature change information.

[0116] As Figure 3 shown, the feature change information includes multiple sub-change information. Therefore, in the specific implementation, for each sub-change information, the server can perform a power operation on the sub-change information and the divergence index to obtain the optimized sub-change information, where the reciprocal of the sub-change information is the exponent and the divergence index is the base. The corresponding mathematical expression is:

[0117] y = x n

[0118] In the formula, y represents the optimized sub-change information, x represents the divergence index, and n represents the reciprocal of the sub-change information.

[0119] In this embodiment, the divergence index can be the ratio or difference between the first risk score and the second risk score. When the divergence index is the difference between the two, the divergence index is also a value within the range of 0 to 1. Through the above expression, if the sub-change information is less than 1, it will be further reduced, and if the sub-change information is greater than 1, it will be further enhanced. Moreover, restricting the value range of the optimized sub-change information to between 0 and 1 is beneficial for the subsequent processing of the weight model.

[0120] In this way, the problem of different value ranges of different risk sub-feature values can be effectively solved, ensuring that all feature change information is compared and processed on the same scale. Secondly, using the divergence index for power operation optimization can flexibly amplify or reduce the amplitude of the sub-change information, thereby more accurately distinguishing between sudden information and non-sudden information. Especially for the case where the sub-change information is less than 1, its influence will be further reduced after optimization, while for the case where the sub-change information is greater than 1, its influence will be further enhanced, enabling the model to more sensitively capture the key change points.

[0121] Based on the above description of the optimization method for feature change information in the embodiment, continue to refer to Figure 2 or Figure 4 , after step S4-2, step S4 further includes:

[0122] S4-3. Process the optimized feature change information through a pre-trained weight model to obtain the respective prediction weights of the first risk score and the second risk score.

[0123] S4-4. Obtain a comprehensive risk score based on the respective prediction weights of the first risk score and the second risk score.

[0124] In this way, the weight model determines the importance of each risk score in the overall assessment by processing the optimized feature change information. After training, the weight model can more sensitively capture key change points, automatically adjust the weights of each risk score, and more accurately reflect its predictive ability for future risks. In addition, the weight model can be a pre-trained machine learning model, such as a neural network or a regression model, which can automatically learn and adjust the weights of each risk score according to the optimized feature change information. Specifically, the model can adopt the method of supervised learning and be trained with a large amount of data so that the model can identify which feature change information is more important for predicting the first risk score and the second risk score. During the training process, the model continuously adjusts its internal parameters to minimize the prediction error and finally obtains a model that can predict the prediction weights of each risk score.

[0125] As introduced in the above embodiments, when the difference between the first risk score and the second risk score is large, it means that there is a large divergence in the recognition results of the two models, which is often caused by the insensitivity of the ARIMA model to sudden information. Therefore, it is not necessary to use the above weight model to adaptively generate the prediction weights of the first risk score and the second risk score every time, but it is necessary to rely on the divergence index between the first risk score and the second risk. Therefore, as Figure 5 shown, on the basis of Figure 4 , Figure 4 step S4-2 in

[0126] also includes: Figure 5 S4-2-1. Judge whether the divergence index is greater than the divergence threshold. If so, execute Figure 5 step 4-2-2 in

[0127] ; otherwise, execute

[0128] sub-step S4-5 of step S4 in

[0129] S4-5. Obtain a comprehensive risk score based on the respective preset weights of the first risk score and the second risk score.

[0128] It can be understood that the server will calculate the divergence index between the first risk score and the second risk score and compare it with a preset divergence threshold. If the divergence index does not exceed the threshold, it is considered that the results of the two models are acceptable. At this time, the respective preset weights can be directly used to calculate the comprehensive risk score. In this way, not only can inaccurate weight allocation be avoided in high-divergence situations, but also the calculation process can be simplified in low-divergence situations, improving the efficiency and robustness of the overall system.

[0129] For the above risk assessment model and autoregressive moving average model, this embodiment proposes a new model training method, which can greatly improve the training efficiency of the model. Taking the risk assessment model as an example, the training method of the risk assessment model will be described in detail below. Before the description, it should be understood that hyperparameters are crucial in model training, determining the learning process and final performance of the model. Appropriate hyperparameter settings can control the complexity of the model, prevent overfitting or underfitting, thereby ensuring the model learns effectively; at the same time, it affects the convergence speed of the model, and the training process can be accelerated through reasonable selection; in addition, hyperparameters also affect the choice and efficiency of the optimization algorithm, and thus indirectly affect the generalization ability of the model.

[0130] In view of this, the server can also call a third-party large language model to generate the current hyperparameters for the network model to be trained; initialize the network model to be trained with the current hyperparameters; train the initialized network model to be trained with training samples to obtain a candidate model trained by the network model to be trained and the training effect; continue to call the third-party large language model to process the current hyperparameters and the training effect to obtain new hyperparameters; according to the new hyperparameters, return to the step of initializing the network model to be trained with the current hyperparameters, and until the iteration stop condition is met, select the one with the best training effect from multiple candidate models as the risk assessment model.

[0131] It can be understood that in the above training method, the server uses the reasoning ability of the third-party large language model to generate hyperparameters for the network model to be trained, and adjusts the hyperparameters according to the feedback training effect, so as to continuously optimize the set hyperparameters.

[0132] Exemplarily, assume that the third-party large language model is Qwen72b and the network model to be trained is Qwen14b. In the early stage of training, a series of hyperparameter limit conditions are preset to optimize the training effect of the model. First, take {initial weight / 500, initial weight * 500} as its initial weight range, and set the learning rate in the interval of [0.001, 0.1], and adjust it with a step size of 0.001, and set the batch size in the interval of [16, 128], with a step size of 16; finally, set the number of hidden layer nodes according to the model complexity, in the range of [32, 256], with a step size of 32.

[0133] Based on the above constraints, the third-party large language model generates hyperparameters within the scope of the above constraints. In order to achieve this process, the server calls the third-party large language model Qwen72b to generate a set of initial hyperparameters for the network model to be trained Qwen14b. These hyperparameters are used to initialize the Qwen14b model. Then, the initialized Qwen14b model is trained using the training sample to obtain a candidate model and its training effect. Subsequently, the server calls Qwen72b again, taking the current hyperparameters and training effect as input, so that Qwen72b performs reasoning based on the called context and the current hyperparameters and training effect to obtain new hyperparameters. According to the new hyperparameters, the server reinitializes the Qwen14b model and repeats the above training and hyperparameter adjustment process until the iteration stop condition is met (for example, the loss value of the candidate model approaches stability). Throughout the process, the server uses the reasoning ability of Qwen72b and the feedback training effect to adjust the hyperparameters and gradually optimize the model. Finally, the one with the best training effect is selected from multiple candidate models as the final risk assessment model. In this way, the scheme effectively utilizes the reasoning capability of Qwen72b and achieves efficient training and optimization of the Qwen14b model.

[0134] The ARIMA model can be optimized in the same way, which will not be described in detail in this embodiment.

[0135] Based on the same inventive concept as the auto loan risk prediction method provided in this embodiment, this embodiment also provides an auto loan risk prediction device, which includes at least one software function module that can be stored in a memory or fixed in an electronic device in the form of software. The processor in the electronic device is used to execute the executable module stored in the memory. For example, the software function module and computer program included in the device. Please refer to Figure 6 , functionally speaking, the device may include:

[0136] The risk characteristic module 11 is used to obtain the overdue risk characteristics of the borrower in the current repayment cycle;

[0137] The risk scoring module is used to process the overdue risk characteristics through the risk assessment model of the neural network architecture to obtain the first risk score of the borrower in the current repayment cycle;

[0138] The risk scoring module is also used to process the overdue risk characteristics through a pre-fitted autoregressive moving average model to obtain the second risk score of the borrower in the current repayment cycle;

[0139] The risk synthesis module 13 is used to obtain a comprehensive risk score of the borrower according to the first risk score and the second risk score.

[0140] In this embodiment, the risk feature module 11 is used to implement Figure 1 step S1 in Figure 1 step S2 and S3 in Figure 1 step S4 in. Therefore, for the detailed description of each of the above modules, reference may be made to the specific implementation manners of the corresponding steps, and this embodiment will not elaborate thereon.

[0141] Since it has the same concept as the above vehicle loan risk prediction method, the vehicle loan risk prediction device can also implement other steps or sub-steps of the method through the above modules.

[0142] Optionally, the risk integration module 13 is further specifically configured to:

[0143] Obtain the feature change information between the overdue risk features of the previous repayment period and the overdue risk features of the current repayment period;

[0144] Optimize the feature change information to obtain the optimized feature change information, where the optimized feature change information enhances the feature change information including sudden information and suppresses the non-sudden information in the feature change information;

[0145] Process the optimized feature change information through a pre-trained weight model to obtain the prediction weights of the first risk score and the second risk score respectively;

[0146] Obtain the comprehensive risk score according to the prediction weights of the first risk score and the second risk score respectively.

[0147] Optionally, the risk integration module 13 is further specifically configured to:

[0148] Obtain the divergence index between the risk assessment model and the autoregressive moving average model according to the first risk score and the second risk score;

[0149] Optimize the feature change information by using the divergence index to obtain the optimized feature change information.

[0150] Optionally, the feature change information includes multiple change sub-information, and the optimized feature change information includes multiple optimized change sub-information; the risk integration module 13 is further specifically configured to:

[0151] For each change sub-information, perform a power operation on the change sub-information and the divergence index to obtain the optimized change sub-information, where the reciprocal of the change sub-information is the exponent and the divergence index is the base;

[0152] Optionally, after obtaining the divergence index between the risk assessment model and the autoregressive moving average model based on the first risk score and the second risk score, the risk integration module 13 is further specifically configured to:

[0153] Determine whether the divergence index is greater than the divergence threshold;

[0154] If it is greater than the divergence threshold, perform the step of optimizing the feature change information using the divergence index to obtain the optimized feature change information.

[0155] Optionally, the overdue risk features of the current repayment period include multiple overdue sub-features and the preset weights assigned to each overdue sub-feature.

[0156] Optionally, the vehicle loan risk prediction device further includes a model training module, and the model training module is used for:

[0157] Invoke a third-party large language model to generate the current hyperparameters for the network model to be trained;

[0158] Initialize the network model to be trained using the current hyperparameters;

[0159] Train the initialized network model to be trained using training samples to obtain a candidate model trained by the network model to be trained and the training effect;

[0160] Continue to invoke the third-party large language model to process the current hyperparameters and the training effect to obtain new hyperparameters;

[0161] According to the new hyperparameters, return to the step of initializing the network model to be trained using the current hyperparameters, and select the one with the best training effect from multiple candidate models as the risk assessment model until the iteration stop condition is met.

[0162] In addition, each functional module in various embodiments of the present application may be integrated together to form an independent part, or each module may exist alone, or two or more modules may be integrated to form an independent part.

[0163] It should also be understood that if the above implementation manner is implemented in the form of a software functional module and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the existing technology, or a part of this technical solution, may be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application.

[0164] Therefore, this embodiment also provides a storage medium, which is a computer-readable storage medium. The storage medium stores a computer program, and when the computer program is executed by a processor, the vehicle loan risk prediction method provided by this embodiment is implemented. Among them, the storage medium can be various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc.

[0165] An electronic device for implementing the vehicle loan risk prediction method provided by this embodiment. As Figure 7 shown, the electronic device may include a processor 22 and a memory 21. And, the memory 21 stores a computer program, and the processor reads and executes the computer program corresponding to the above embodiments in the memory 21 to implement the vehicle loan risk prediction method provided by this embodiment.

[0166] Continue to refer to Figure 7 , the electronic device further includes a communication unit 23. The memory 21, the processor 22, and the communication unit 23 are electrically connected to each other directly or indirectly through a system bus 24 to achieve data transmission or interaction.

[0167] Among them, the memory 21 can be an information recording device based on any electronic, magnetic, optical, or other physical principles for recording execution instructions, data, etc. In some embodiments, the memory 21 can be, but is not limited to, a volatile memory, a non-volatile memory, a storage drive, etc.

[0168] In some embodiments, the volatile memory can be a random access memory (Random Access Memory, RAM); in some embodiments, the non-volatile memory can be a read-only memory (Read Only Memory, ROM), a programmable read-only memory (Programmable Read-Only Memory, PROM), an erasable programmable read-only memory (Erasable Programmable Read-Only Memory, EPROM), an electrically erasable programmable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), a flash memory, etc.; in some embodiments, the storage drive can be a disk drive, a solid-state drive, any type of storage disk (such as an optical disc, a DVD, etc.), or a similar storage medium, or a combination thereof, etc.

[0169] The communication unit 23 is used to transmit and receive data via a network. In some embodiments, the network may include a wired network, a wireless network, an optical fiber network, a telecommunication network, an intranet, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), a wide area network (WAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigBee network, or a near field communication (NFC) network, etc., or any combination thereof. In some embodiments, the network may include one or more network access points. For example, the network may include a wired or wireless network access point, such as a base station and / or a network switching node, and one or more components of the service request processing system may be connected to the network via the access point to exchange data and / or information.

[0170] The processor 22 may be an integrated circuit chip with signal processing capabilities, and the processor may include one or more processing cores (e.g., a single-core processor or a multi-core processor). By way of example only, the above-mentioned processor may include a central processing unit (CPU), an application specific integrated circuit (ASIC), an application specific instruction-set processor (ASIP), a graphics processing unit (GPU), a physics processing unit (PPU), a digital signal processor (DSP), a field programmable gate array (FPGA), a programmable logic device (PLD), a controller, a microcontroller unit, a reduced instruction set computing (RISC), or a microprocessor, etc., or any combination thereof.

[0171] It can be understood that Figure 7The structure shown is only illustrative. The electronic device may also have more or fewer components than Figure 7 shown, or have a different configuration from Figure 7 shown. Figure 7 Each of the components shown may be implemented by hardware, software, or a combination thereof.

[0172] It should be understood that the devices and methods disclosed in the above embodiments may also be implemented in other ways. The device embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, may be implemented by a dedicated hardware-based system for performing the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

[0173] As described above, these are only various embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for predicting car loan risk, characterized in that: The method comprises: Obtain the borrower's overdue risk characteristics in the current repayment cycle; Processing the overdue risk characteristics through a risk assessment model with a neural network architecture to obtain a first risk score of the borrower in the current repayment cycle; Processing the overdue risk characteristics through a pre-fitted autoregressive moving average model to obtain a second risk score of the borrower in the current repayment cycle; A comprehensive risk score of the borrower is obtained based on the first risk score and the second risk score.

2. The method for predicting auto loan risk according to claim 1, characterized in that: Obtaining a comprehensive risk score of the borrower according to the first risk score and the second risk score includes: Obtaining feature change information between the overdue risk features of the previous repayment cycle and the overdue risk features of the current repayment cycle; Optimizing the characteristic change information to obtain optimized characteristic change information, wherein the optimized characteristic change information enhances the characteristic change information including burst information and suppresses non-burst information in the characteristic change information; Processing the optimized feature change information through a pre-trained weight model to obtain respective prediction weights of the first risk score and the second risk score; The comprehensive risk score is obtained according to the prediction weights of the first risk score and the second risk score.

3. The method for predicting auto loan risk according to claim 2, characterized in that: Optimizing the feature change information to obtain optimized feature change information includes: Obtaining a divergence index between the risk assessment model and the autoregressive moving average model according to the first risk score and the second risk score; The feature change information is optimized using the divergence index to obtain optimized feature change information.

4. The method for predicting auto loan risk according to claim 3, characterized in that: The feature change information includes a plurality of change sub-information, and the optimized feature change information includes a plurality of optimized change sub-information; The feature change information is optimized using the divergence index to obtain optimized feature change information, including: For each piece of the change sub-information, a power operation is performed on the change sub-information and the divergence index to obtain optimized change sub-information, wherein the reciprocal of the change sub-information is the exponent and the divergence index is the base.

5. The method for predicting auto loan risk according to claim 3, characterized in that: After obtaining a divergence index between the risk assessment model and the autoregressive moving average model according to the first risk score and the second risk score, the method further includes: Determining whether the divergence index is greater than a divergence threshold; If it is greater than the divergence threshold, the step of optimizing the feature change information using the divergence index to obtain optimized feature change information is performed.

6. The method for predicting auto loan risk according to claim 1, characterized in that: The overdue risk characteristics of the current repayment cycle include multiple overdue sub-characteristics and a preset weight assigned to each of the overdue sub-characteristics.

7. The method for predicting auto loan risk according to claim 1, characterized in that: The method also includes a training method for the risk assessment model, the training method comprising: Call a third-party large language model to generate current hyperparameters for the network model to be trained; Initialize the network model to be trained using the current hyperparameters; The initialized network model to be trained is trained by using training samples to obtain a candidate model trained by the network model to be trained and a training effect; Continue to call the third-party large language model to process the current hyperparameters and the training effect to obtain new hyperparameters; According to the new hyperparameters, the step of initializing the network model to be trained using the current hyperparameters is returned to until the iteration stop condition is met, and the model with the best training effect is selected from the multiple candidate models as the risk assessment model.

8. A car loan risk prediction device, characterized in that: The device comprises: The risk characteristics module is used to obtain the overdue risk characteristics of the borrower in the current repayment cycle; A risk scoring module, used to process the overdue risk characteristics through a risk assessment model with a neural network architecture to obtain a first risk score of the borrower in the current repayment cycle; The risk scoring module is further used to process the overdue risk characteristics through a pre-fitted autoregressive moving average model to obtain a second risk score of the borrower in the current repayment cycle; The risk synthesis module is used to obtain the comprehensive risk score of the borrower according to the first risk score and the second risk score.

9. A storage medium, characterized in that: The storage medium stores a computer program, and the computer program, when processed by a processor, implements the auto loan risk prediction method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores a computer program, and the computer program is processed by the processor to implement the auto loan risk prediction method according to any one of claims 1 to 7.

Images